Herbert Wang retweeted
Claude's new tools just killed Microsoft Office.
So we open-sourced our own versions to kill Claude's.
Whiteboard. Spreadsheet. Dashboard.
Free. Open source. Works with any model. Available now.
Folders organize work for humans. For agents, they define what they can understand and act on.
PuppyOne is file-level context management for local workspaces—with Codex and Claude Code in the same editor.
📁 Notion
┃
┣ 📁 System of Record
┃ ┣ 📁Databases
┃ ┣ 📁Meeting Notes
┃ ┣ 📁Version Control
┃ ┣ 📁Wiki
┃ ┣ 📁Projects
┃ ┣ 📁Docs
┃ ┣ 📁Tasks
┃ ┣ 📁Forms
┃ ┗ 📁Knowledge base
┃
┣ 📁 Notion AI
┃ ┣ 📁Custom Agents
┃ ┣ 📁External Agents
┃ ┣ 📁Enterprise Search
┃ ┣ 📁AI Meeting Notes
┃ ┣ 📁Deep Research
┃ ┣ 📁Autofill
┃ ┣ 📁AI Connectors
┃ ┗ 📁Mobile Agents
┃
┣ 📁 Developer Platform
┃ ┣ 📁MCP
┃ ┣ 📁Workers
┃ ┣ 📁API
┃ ┣ 📁ntn CLI
┃ ┣ 📁Webhooks
┃ ┣ 📁Syncs
┃ ┣ 📁Agent tools
┃ ┣ 📁Developer Portal
┃ ┗ 📁External Agents API
┃
┣ 📁 Connected Workspace
┃ ┣ 📁Slack
┃ ┣ 📁Calendar
┃ ┣ 📁Mail
┃ ┣ 📁GitHub
┃ ┣ 📁Jira
┃ ┣ 📁Asana
┃ ┣ 📁Figma
┃ ┣ 📁Zapier
┃ ┣ 📁Salesforce
┃ ┗ 📁Google Drive
┃
┣ 📁 Business Workflows
┃ ┣ 📁CRM
┃ ┣ 📁Roadmaps
┃ ┣ 📁Launch plans
┃ ┣ 📁Editorial calendars
┃ ┣ 📁Customer research
┃ ┣ 📁Team updates
┃ ┣ 📁Onboarding
┃ ┣ 📁Approvals
┃ ┗ 📁Reporting
┃
┣ 📁 Personal Productivity
┃ ┣ 📁Personal OS
┃ ┣ 📁Notes
┃ ┣ 📁Tasks
┃ ┣ 📁Goals
┃ ┣ 📁Reading lists
┃ ┣ 📁Travel planning
┃ ┣ 📁Habit tracking
┃ ┣ 📁Side projects
┃ ┗ 📁Life admin
54
One lesson from running AI workflows every day:
The model is rarely the bottleneck.
Keeping context synchronized across Claude Code, Codex, Cursor, OpenClaw, and other runtimes is much harder than most people expect.
A good write-up from our founder on why we built puppyone.
3
107
“Context infrastructure” is going to matter more than people think.
Multi-agent workflows need shared memory, scoped permissions, and reproducible state.
After months of shipping, we launched @puppyone_ai 3.0: context infrastructure for multi-agent teams.
Most agent teams don’t fail at models first.
They fail when agents don’t share a single source of truth for context.
What this means in practice:
1️⃣SaaS context is materialized into agent-readable files
2️⃣Each runtime gets scoped access (least privilege)
3️⃣Every write is versioned, auditable, and rollback-ready
So your research agent, coding agent, and reviewer agent can work from the same governed workspace - without sharing human admin credentials.
If you’re running OpenClaw, Claude Code, Codex, or Hermes, puppyone is designed to be the context layer across MCP, CLI, and API.
Try: try.puppyone.ai
CLI: puppyone.ai/en/cli
Repo: github.com/puppyone-ai/puppy…
Full launch post——See more ⬇️
#AIEngineering #AIAgents #MCP #DevTools #OpenSource #puppyone30 #context
14
Herbert Wang retweeted
Did I mention that we're launching a 'context folder' for agents? It lets you gather all the context in one place! 😆
Building an AI agent demo feels like building a startup. Until you try to sell it.
I’m in that phase right now.
At what point does a demo become something people actually pay for?
23
Trying OpenClaw.
Spent more time worrying about protecting my local data than actually using the agent.
For non-devs, this is… a lot. @openclaw
Is it just me?
1
28
If people truly trusted agents,
there wouldn’t be so many Mac minis running OpenClaw.
Agent capability isn’t the problem.
Trust is.
2
28
Perplexity is good, but not enough.
I want research that expands the way I need it to: when to go deeper, when to go wider, which sources to trust, and how much to spend.
Open Deep Wide Research 🆚 Perplexity
Same question — two different outcomes.
We’re not building a ‘better search.’ We’re building a controllable, private, deep-reasoning research agent.
You control:
· reasoning depth
· exploration width
· cost
· data sources
1
1
58
It’s the summary of what we’ve learned so far,
and the reason we started building this in the first place.
We built Open Deep Wide Research on a simple idea:
We should control over the depth and breadth of Agentic RAG.
medium.com/@realguantum/rag-…
39
Herbert Wang retweeted
We’re Open-Sourcing: Open Deep Wide Research:
The first Agentic RAG system that lets you precisely customize the trade-off between quality, response time and cost for any RAG scenario
We give you full control over the depth and breadth of Agentic RAG, as well as data sources, to precisely match your cost, response time, and quality requirements.
Shipping a new open-source product next week.
Now thinking about the hardest part: the cold start.
Any tips or playbooks you’d recommend to get the first wave of users?
2
11,191
Low-code was born to simplify software. Manually breaking work into predictable functions and wiring them together.
But work has changed — now about information and context. Large language models handling the messy process in between.
A paradigm shift is coming for no-code.
Our PuppyAgent 101 playlist is really starting to stack up 🎬Guess we’re basically running a mini-course now.
Haven’t started yet? Today’s a good day to jump in👇
youtu.be/Sz6oh4U-wmE?si=HNM3…
79
Exactly.
Right now, even using AI well requires a real skill set — let alone replacing human work.
With PuppyAgent, our goal is to make AI truly useful for the kind of paperwork and routine tasks everyone deals with.
Never be afraid your job will be replaced by AI.
Unless your job is just copy-paste from ChatGPT.
Then be afraid.
Be very afraid.
43
This is the real force of a great product.
That’s exactly the kind of impact we’re striving for——aiming to build something that powerful.
🍌 Nano Banana: The AI spark igniting 10M+ Gemini users in a week! Like the Ghibli Art trend, it’s reshaping creativity with photorealistic magic. Will it outlast the hype? Try it! @NanoBanana @GeminiApp
1
26
















